The "Undervolting" Guide

I just found out about this website and your undervolting guide, and since my laptop gets really hot, I decided to go ahead and do it, seeing as there's no catch whatsoever.
I downloaded RMclock, and tried to follow your guide, but the Advanced CPU Settings tab isn't there, and neither is the Performance on Demand tab. Any reason why? I have 64-bit Vista, but I installed the correct driver.
If it matters, my CPU is an AMD Turion X2 Dual-Core Mobile RM-72.

 

I have been trying to use this guide to undervolt my alienware laptop to reduce some heat. It has an intel T5500 and at maximum power CPU-Z says it's using about 1.288V. I figured I could probably end up cutting down on that a lot.

I've ran into 2 problems though and having looked through quite a few pages to see if anybody else has ran into the same problems. First of all the highest voltage it will allow me to go up to is 1.1750 on the program so that might be to big of a jump to drop volts down instantly from the 1.288V.

I tried it anyways to see if it would work and followed the setup on the first page but no matter what I set my 10X multiplier to, it doesn't end up changing the value (at least according to CPU-Z). It still report it at 1.288V. I've checked all the indexes and made sure it's on "performance on demand". The CPU info on RMclock shows that at 10X multiplier says Req Voltage is 1.1750 but I assume that's no telling me what it is actually at.

Would the bios be stopping me from altering the voltages?

 

Thank you, I tried the undervolting and the temperature of my laptop cpu dropped about 10 degree.

 

I could really use some input on how to undervolt processors with half multipliers without loosing any performance.

I'm currently using a p8400. I've tried Crystal CPUID, but for some reason none of my custom settings takes effect.

Any help would be greatly appreciated!

 

RMClock will not work for the newer AMD cpu's. Try this one instead...
http://amd.goexchange.de/

 

Not just HP. All of them should undervolt during their QC stage before shipping. It's one thing to put in an under-powered cooling system to get "points" from the quickie reviewers for low noise and longer battery life. It's another thing that Intel introduced speed switching for low power mobile CPUs when it's a complete waste of time (literally !). Can ANYBODY tell me why speed switching is necessary if the CPU doesn't overheat ? Bonus question: When most people are using 2 or 3 GB of RAM, why does Windows need to page stuff back out to the hard drive where it came from in the first place

A box of a dozen doughnuts for the first GOOD answer. Judges are standing by.......

 

Okay, I downloaded that, and installed that, and I'm stuck...
I have NO idea how to use Griffin Control, as the interface is command prompt. Any instructions/pointers/advise/any help at all??

 

In response to my last post, I found out that the voltages were changing even though they do not correspond to the values listed on RM clock. what surprised me is that I went from 1.288V at stock 1.66Ghz speeds to 1.063V at the same speeds. My power saving mode runs the cpu at those volts. It is still fully functional at .225 volts below what it was set previously and the only reason I couldn't get any lower is because the program didn't give me any lower voltage options.

Anything I can do show more voltage options? If not then I may end up trying to overclock the thing, except last time I tried clockgen wouldn't let me.

 

Ok, i dont know if a guide exists. But here's the basic outline (read through it completely before trying):

1. Download both the "Supporting file" and the "Windows service installer" from this page...
http://amd.goexchange.de/#downloads

2. The "Supporting file" download should be in the form of a compressed zip file. Extract the file using winzip, 7-zip etc and put it into a folder of your choice.

3. Drag the Windows service installer exe to the same folder as in step 2.

4. Run the GriffinControlservice.exe and type 'i' (without quotes) and give the service a name.

5. Exit out of the command window. The service has now been created.

6. Look in the folder you created in step 2, you should see a XML document named GriffinControl_config. Open this file in Notepad.

7. At the very top of the file you should see this...

PHP:

<pumastatectrl use_p_state_0_as_max_freq="true" log_file_path="" confirm="false">
  <p_states>
   <pstate number="0" Vid='41'></pstate>
   <pstate number="1" Vid='60'></pstate>
   <pstate number="2" Vid='64'></pstate>
  </p_states>

NOTE: Do not change anything else in this file other than the VID numbers.

Each pstate number you see above refers to the different states possible with your cpu. For the RM 72, pstate 0 = 2.1GHz, pstate 1 = 1.1GHz and pstate 2 = 0.5GHz. The VID's refer to the corresponding voltages. The voltages for thecorresponding VID's can be found here...
http://amd.goexchange.de/configuration_file.html

Set the appropriate voltages (i would suggest you start off with the stock VID/voltages and slowly work your way down testing at each VID/voltage and pstate.

8. Finally, go to Start -> in the search box type services.msc and hit enter-> You should then see a list of services and in that list should be the one you created in step 5. Start the service and your VID/voltages should have been applied. You can then proceed to check for stability.

NOTE: Set it to run automatically only after you have worked out stable voltages.

 

I also have a P8400, and i undervolted it. If you dont use CPU-heavy programs (video editing etc), you wont notice the 133MHz difference, and it really makes your laptop cooler.
I think being cooler outweighs the slightly reduced Freq. during normal usage.

 

Click on the thumbnails below. Use your own highest index number for your 8400. Don't check any others. Send Lefsa, keep the Lutefisk.

 

I did all that and it worked!! I ran an ORTHOS CPU test before and after, and before the CPU temp got up to 90C, and now it only got up to 80C! That's a decrease of 10 degrees Celcius, even though I was only able to lower the voltage from 1.05 to 1.037!
One thing still worries me however. I downloaded CPU-Z to see if it the voltage change was in order, and I noticed that the Core VID keeps changing back and forth from 1.037 V (what it's currently set to) to 1.200 V. I looked on the VID guide on the GriffinControl website, and 1.200 isn't even on there. leading me to believe that 1.200 V is BAD. I certainly hope that CPU-Z is just acting up, because I'm kinda freaking out right now... That might, however, explain why the CPU still gets up to 80C, which according to my understanding is still pretty hot...
Any ideas??

UPDATE: I also noticed that in the Clocks box in CPU-Z that the Multiplier keeps changing back and forth from x5.25 to x10.5, causing the core speed to double while it's momentarily at x10.5. Is there any correspondence between that and the voltage fluctuation? I am very confused...

SECOND UPDATE: After about a half hour of intense gaming, it BSOD'd, so I disabled GriffinControl, and ran CPU-Z afterwards, and it turns out I decreased the voltage by waaaay more than I thought I did. My mistake was thinking that the placeholder VIDs in the GriffinControl_Config folder were my actual VIDs. I thought it was at 1.05 V before, but it was really at 1.125 V, which seems a bit high to me. Also, even without GriffinControl enabled, it still fluctuates between 1.125 V and 1.200 V. I'm gonna re-do the undervolting later today, and see what the ACTUAL lowest stable voltage is. I don't think that'll stop it from fluctuating between the set voltage and 1.200 V though, so I'm still REALLY confused...

 

i dont understand what you mean

 

I too am intrigued. Send Lefsa, keep the Lutefisk? But fish is good for you!

I mean I'm also intrigued as to why you'd set your multiplier on max all of the time; which is what you are suggesting am I correct?

 

I too, am a Norwegian Herring Choker (gill net). All of us are out here in the fjords of Seattle. Minnesota has emptied out by now. I love fish, fried, not turned into Jello with lye !

OK, enough introductions. Now business. If your notebook has been undervolted correctly it will NOT overheat at full CPU clock frequency and 100% load. Right ??? And, battery drain depends only on how much you actually use your computer. Standby takes about one tiny little watt of power. Idle, with screen blanked and HDD off takes about 5 watts.

OK, I'm turning the tables here. For a hunk of Puget Sound Smoked Salmon, GIVE ME ONE GOOD REASON WHY YOU NEED TO RUN YOUR NOTEBOOK SLOWER THAN IT CAN GO......

Used to be Scott Agerskov. Thanks Grandpa

 

Alright, so you're basically saying that there is absolutely no need for the speedstepping feature?

I'll be giving it a try in the morning, thank you for the response.

 

The change in the multiplier values relates to the cpu clock speed and its perfectly normal for it to change along with the clock speed. As far as the 1.2V spike is concerned, it happens on my ZM 86 as well. Why it happens is a mystery to me and from what I can remember it happened even when not undervolted. Your temps are quite good. Max of around 80C is not too bad at all particularly for this generation of Turions. BTW, the stock voltage should be around 1.075V. 1.1V is the max default for the ZM's.

 

Undervolting is a necessity. Speed-stepping is an amusement, IMHO.

 

Trying to remember his physics lessons all those years ago...... v=IR etc

The larger the wattage the more heat is given off. Lower resistance for a given voltage equals more wattage which as validated by the first statememt means more heat production.

If you crank the voltage up the wattage goes up, also if you crank up voltage without changing resistance the amperage goes up as well, which also increases wattage. Remember that increasing wattage increases heat.

As the frequency of the processor decreases the resistance increases which means lower amperage and lower wattage for a given voltage.

Therefore, and this is assuming I've got all of the above correct, having the processor run at a lower clock speed at the same voltage means lower temperature.

Now the question is, does changing your multiplier increase resistance?

I think it must. The reason I say this is that as part of the speedstep technology (according to this PDF on page 19) the processor will drop down to a lower multiplier if the temperature reaches a certain level. Why would it do that? Well the way I see it, it must be that changing the multipler does increase resistance also.

I may be waaaaay of the mark here; it has been many many many years since I've exercised my brain in this area. I am ready and willing for all and sundry to advise me that I am wrong.

 

"As the frequency of the processor decreases the resistance increases which means lower amperage and lower wattage for a given voltage"

You were doing terrific until you crashed into the statement above. Consider a semiconductor as more or less a fixed resistance like a light bulb. You're right about raising the voltage causing an increase in current (amps) AND watts because watts = volts X amps. Also, in a resistor, all of the watts consumed are given off as waste heat. Conversely, lowering voltage across a fixed resistance lowers current flow (amps) and lowers watts consumed and released as heat (sounds like undervolting, heh heh).

Here's the tricky part. Think of a transistor as a switch. It is. A tiny voltage change can start or stop a very large flow of current i.e. electrons. It's the massive flow of current through the natural resistance of the circuits that creates friction and generates heat.

OK. Now, think about a light switch on a lamp. I don't care how fast or slow you flip the switch on and off; if there's no bulb in the lamp then there's no current flow, no watts used, no heat produced. Consider your CPU as a variable light bulb. When it's busy (load) it's shoving current all over the place creating heat. When it's quiet, at no load, few electrons are moving and little heat is produced. Doubling the oscillator frequency sends current around the circuit twice as often, NOT twice as fast. (electrons have speed limits too). So, at a fixed load and a fixed voltage, doubling the frequency does double the watts used.

Here's the Bingo that went off over my head: A computer program is fixed length set of instructions. If you double the frequency, you double the total current flow over only half the normal time interval. Then you're done. You finished the instruction set in half the regular time. It's like comparing two workers. One works twice as fast and is done in half the time. The man-hours expended is the same for both. Your electric is bill is in Kilowatt-hours and your laptop battery is rated in milliamp-hours. (mW-hrs = mA-hrs X battery voltage). Consider the job time !!!

Me, I'll take twice as fast and half the time as long as I don't overheat...... Undervolting is important because it lets you run fast without burning out. Speed-stepping is just a slow-down safety valve for under-cooled stock CPUs belonging to those unfortunates who failed to follow the teachings of Rev. Flipfire....

Somebody owes me a pint of bitter next time I venture across the Pond, heh heh.

 

So, is the voltage spike bad? And if so, is there a way to fix it?
Also, I noticed that your HP Pavilion has a Radeon HD 3200, the same card I currently have. Have you had any problems with it, specifically with games? Whenever I play games for a while, and the temperature inside my computer approaches 80, I get weird graphical glitches no matter what game I'm playing. I know that's a bit off topic, but have you had similar problems? Or would you know what causes that?

 

OK Scott thanks for the explanation. I think I'll have to take some time to get my head around this, but that's not right now. I expect I'll be posting further questions on this though, those A-levels really were longer ago than I remember.

 

Unfortunately, HP these days seem to have dropped the ball when it comes to cool running notebooks. Many of their systems run hotter than what other manufacturers offer. I have a 12" system which means there isnt enough space in the first place food good cooling systems and so mine plus others who own the same system quite regularly see cpu temperatures max out around 100C, with undervolting I have been able to bring it down a bit but unfortunately with HP's inadequate cooling the cpu still does run quite high. I havent seen any glitches even at 100C, what driver version are you using. I would try to get the latest catalyst driver from the AMD website and see if it helps.

 

Thanks for the nice guide!

I followed the instructions thoroughly but my laptop freezes whenever SuperLFM is enabled. The moment I enable it, the whole system halts. It's rather weird, really.

The laptop is a Dell XPS Studio 13 with a P9600. I've disabled both IDA and most of the P-states. Only P1 - 6x 0.9250V and P5 - 10x 1.1V are enabled right now. I'd love to get SuperLFM to work, any hints?

P.S. : All drivers (including BIOS) are the latest available at Dell.com

 

Awesome post, flipfire!

Attempting this with my Dell XPS M1330, P9300 Penryn 2.5Ghz. Running Win 2008 Server Standard 64-bit.

Unfortunatlely I am stuck. I have set up RMClock as you have suggested (checked and double-checked!), but when I run ORTHOS, RMClock and CPU-Z report Core clock of about 1.2Ghz with a multiplier of 6.0x (instead of the 12.0x I expected). RMClock reports a Throttle at c.1196MHz

RMClock profile is set as per your example (except I have a 12x multiplier available). I have not touched IDA and LFM as per your guidance. I have not even changed the voltages yet!

Okay, I have rebooted and started ORTHOS without RMClock. CPU-Z shows the higher multipliers being used and HWMonitor shows temps. increasing. As soon as I start RMClock, the CPU is throttled back to 1.2GHz and 6x multiplier. If I shut down RMClock the hardware remains throttled at 1.2GHz/6x. RMClock Throttling ((Advanced CPU Settings) is 'Inactive'.

Next thing: I disabled all p-states except FID 12x in RMClock and re-run ORTHOS. CPU-Z and RMClock *still* report 6x multiplier as being active. This is starting to drive me nuts!

What could be stopping the higher clock multipliers from being used? Clearly something is wrong with my RMClock set up, but I cannot see it.

Can anyone help? Thanks

J.

 

Jgrogan - Have you got throttling unchecked in RMC? Something is putting on the brakes. I would suggest getting your voltages down first to avoid any braking mechanism from engaging. Be sure all the power profiles (Windows and RMC) match up. Takes a while to get it just right. A server is THE perfect App for RMC.

 

I'd love to get SuperLFM to work - Why? Not good for anything, IMHO. Besides, I think you have to enable IDA to get SLFM. Not worth it.

 

Thx, Scott - I'll look at these. I know RMC has Throttling set to 'Inactive', so that's out. I'll look into your other suggestions & report back / ask for more help etc.

Quite clearly "I can't brain today...I have the dumb."

Cheers
James

 

Scott, SLFM works fine on my Vostro 1720/P8600 without IDA enabled. Much lower voltage at idle resulting in even lower idle temps. 0.875v SLFM at 6X multi at idle.

 

I'm trying to get some "out of the box" thinking going on. Read my long post a page or so down for details. Here's the gist of it: Your CPU doesn't consume power or generate heat unless it's working on something. That's load. The clock frequency driving the processor has virtually zero impact on power use during true idle. Undervolting assures us that at top clock speed AND 100% load, we don't exceed a reasonable operating temperature. Once you have that, you're home free. All this speed switching and stuff just slows down the execution speed of all your operating system and software not to mention chewing up precious clock cycles figuring out what "gear" you're supposed to be driving in. To paraphrase another member here, "Undervolting is essential. Speed switching is entertaining".

 

Undervolting is great. It helps my lappy cool a lot. But I'm worry about voltage spikes. I changes Minimal CPU Utilization in Vista Power Option to 100%. But voltage spikes still happen. Can these spikes harm my computer?
Please help!

 

If by voltage spikes, you mean the voltage on your processor spiking up to 1.2V, then back to what it's set at, then that's the same problem I have.
A poster on these forums (link at end of post) made a program that can reportedly fix it, but I'm waiting for someone to reply in that thread as to how to work his program.

http://forum.notebookreview.com/showthread.php?t=373878

 

My Core temp is around 37 C. Is that good or bad?

 

Your own core temp is a match. 37C. is 98.6F. Good news: your CPU will outlive you

 

just curious, i am running Orthos after doing customization/undervolt the cpu. after 3 hours, the Orthos will beep and there warning message about torturing the cpu sorta like that, but no blue screen happen yet. so the question is, am i safe with current setting? or should i increase the cpu volt until Orthos doesnt give the beep anymore? thanks.

 

hey, im doing the undervolting. however, i haf one v perplexing prob.
my highest FID is 9.0, and at tt level my stable voltage is 1.0625. my lowest FID, 6.0, has a voltage of 0.9875(by default, cant go lower). my 8.0 FID, at 1.0125, hasnt given me a blue screen. between 1.0125 and 0.9875 there's only 1 value left, 1.0000. the values i get to play around with seem v limited.. any explanations for tt?
and also, how to check for voltage spikes?

 

so has anyone tried undervolting a netbook??

 

I got to this stage, but when I choose mobile and click "apply on startup" it still won't show the information of my CPU.
All I see is this:


Should I just continue to follow the guide or is something wrong


Also - The CPU temps I get when gaming are higher than the max temps I got using orthos, any specific reason for this???

Great guide BTW

 

RMClock is 2 years old. It doesn't know what a P8400 is but it works anyway. Also, Orthos doesn't work out your video card, just the CPU. That's where I presume the extra heat is coming from during gaming.

 

From the RMClock website:

The P8400 is a Core 2 Duo is a Penryn core and hence has not been coded into the application. The last update to RMClock was in February 2008 I believe, and I think the P8400 went to market in May/June of 2008.

The site also says

However, I believe many many people have got it to work with no fuss on a Penryn core; just that some functionality might not be available.

 

So I should be able to undervolt fine even though it's not properly recognized.
And also, if I see drops when testing with orthos after the undervolt, I should also see drops when gaming because the CPU will produce less heat when undervolted???

I'll continue with the guide tomorrow morning
Thanks guys

 

When I am done undervolting, do I need to check SuperFLM and IDA in the profiles tab?

 

Like I said, man many people have got it to work, so you shouldn't have any problems. I believe Scott was referring to the fact that the GPU is used more for gaming than the CPU, and so as that heats up some of the heat will dissipate to other components, such as the CPU. Orthos only stress tests your CPU (and RAM for some of the tests); therefore you wont see such high temperatures.

FYI, typically GPUs get much hotter than CPUs

 

OK thanks man
I'll give it a go in the morning and post my results

 

RMClock 2.35 doesn't recognize the name of my T8100 which is earlier than the P8400. That has not prevented it from working perfectly, at least with XP-SP3. Software built like a Russian T-72 tank

 

Works with my T8300. It's the half multipliers with the new CPUs that trips RMclock if I recall correctly.

 

Ok, I have a couple of questions please.

1st off I have a Panasonic Toughbook CF-30, running an Intel Dual Core L7500. I ran the Gromas Core test for 15 minutes with my max temp hitting 158 degrees.

My questions are; Isn't these temps pretty low compared to the usual from others I am seeing while running a 100% load test?

Second, why wont it show my voltages? It just shows my CPU, HDD, AND ACPI temps? What am i doing wrong?

BTW, my laptop has no fans, it's all enclosed for protection.

 

Yeah, it can't see half multipliers, so it downclocks to the nearest whole multiplier

e.g. in my P8400 it only sees max multiplier as 8 instead of 8.5, even when its running at 8.5

 

Depends on who's tripping whom I've posted several times that I worked around the "problem" sucessfully. On my T8100, the true top multiplier is 10.5 giving 2.1 GHz. Well, RMClock offers an 11X box (says IDA) so I checked it and then turned off IDA. Now, my rig runs at 2.1 GHz all day long, on both cores, with no spikes. RMClock says 10X on one screen and 11X on another. Don't care. Multiple other utilities agree with the 2.1 speed.

 

I'd love for you to test out my fix on your newer model CPU. On Profiles, check only the last box, IDA. Leave all the others unchecked. Then, on the Max Performance profile, check the "use P-state" box plus the last drop-down FID box. Lastly, uncheck all the boxes for enhanced features as per my thumbnail. If everything goes to plan you should get a nice flat line at 2530 Mhz, for both cores in the monitor box.